Diagnostic Ability of Impulse Oscillometry in Diagnosis of Chronic Obstructive Pulmonary Disease

Abstract

The diagnosis of chronic obstructive pulmonary disease (COPD) mainly relies on spirometry. Due to the complexity of spirometry, easier-to-do impulse oscillometry (IOS) has been introduced as a complementary approach to conventional pulmonary function testing. Therefore, this study aimed to assess the efficacy of the diagnostic ability of IOS for diagnosing chronic obstructive pulmonary disease (COPD). This cross-sectional study was conducted at the Lung Health Center, Chiang Mai University, Thailand, between June 2019 and January 2020. IOS and spirometry were performed with all subjects suspected of having COPD. A Receiver Operating Characteristic (ROC) curve was plotted, the area under the ROC (AuROC) and 95%CI were compared among COPD and chronic smokers. One hundred and seventeen subjects suspected of having COPD with a mean age of 68.6 ± 8.6 years old were enrolled. Of these 103 (88.0%) were male. Thirty healthy subjects were also enrolled. IOS parameters including resistance at 5 Hz (R5), resonant frequency (Fres), area under reactance (AX), heterogeneity of resistance (R5-R20), and reactance at 5 Hz (X5) demonstrated excellent overall accuracy relative to the diagnosis of COPD with an AuROC ranging from 0.80 − 0.84. The AX ≥ 8.66 cmH₂O/L represented an AuROC = 0.79, with a sensitivity of 79.1% and a specificity of 78.0% for the diagnosis of COPD. IOS is a valuable tool for use in the diagnosis of COPD. It may be used in subjects who cannot carry out the spirometric procedure.

Keywords:

• COPD
• spirometry
• airway resistance
• smoking
• impulse oscillometry

Introduction

Chronic obstructive pulmonary disease (COPD) is characterized by an airflow limitation that is not fully reversible [1]. This airflow limitation is progressive and associated with an abnormal inflammatory response of the lungs to toxic particles or gases [1]. This is mainly attributable to the narrowing of the airways caused by wall remodeling and wall collapsibility during expiration due to loss of alveolar attachment and elastic recoil of the small airways [1]. The World Health Organization (WHO) reports that COPD is one of the leading causes of morbidity and mortality worldwide which results in an estimated global three million deaths every year [2].

At present, the diagnosis of COPD relies mainly on spirometry results. It is recommended as the gold standard for diagnosis of COPD [3]. However, spirometry requires close coordination of maximal inspiratory and expiratory efforts. Many subjects can’t perform spirometry easily, especially those with cognitive impairment, poor motor coordination, or breathing difficulties [4]. Previous studies have shown that unacceptable spirometry results were common in clinical practice (inadequate readings ranging from 8.9%-19.8%) [5,6]. Not only have unacceptable spirometry results been cited as a cause of inaccurate diagnosis but there has also been an issue concerning the lack of qualified spirometry technicians. Previous studies have shown that without training, technicians can only achieve acceptable spirometry readings in between 16.0%-19.2% of cases [6,7]. Therefore, impulse oscillometry (IOS), which is much easier to perform, has been introduced as a complementary approach to conventional pulmonary function testing [8,9].

IOS is a simple, noninvasive method that requires minimal subject cooperation, and allows evaluation of lung function through the measurement of both airway resistance and airway reactance [10]. It was developed by Michaelson et al. in 1975 [11]. The IOS parameters represent the total mechanical properties of the respiratory system [12]. However, the latest guidelines have been recommended that the IOS must be measured followed standardization because there are potential sources of error and variation including variations in patient breathing, bacterial filters, artificial airways, device hardware and signal analysis, data processing, and quality control strategies in IOS measurement [12]. In the past few years, evidence describing effective use of IOS as a lung function assessment tool in COPD is growing [9,13–20]. For example previous studies have verified that many IOS parameters, including resistance at 5 Hz (R5), resistance at 20 Hz (R20), heterogeneity of resistance (R5-R20), reactance at 5 Hz (X5), resonant frequency (Fres), and area under reactance curve between 5 Hz and Fres (AX) in COPD subjects, were significantly higher than chronic smokers [9,14,20] and healthy subjects [15,19,20].

Some parameters in IOS may be used to predict COPD [14,15]. For example, R5, R5-R20, X5, Fres, and AX showed an acceptable area under the receiver operating characteristic curve (AuROC) for accurate diagnosis of COPD when compared to healthy subjects (AuROC ranged from 0.829-0.905) [15]. The previous studies also demonstrated the cutoff value of IOS parameters could be used for detection of COPD [14,15]. Moreover, the cutoff value of IOS parameters could be used for prediction of forced expiratory volume in the first second (FEV₁) at < 50% of predicted value in COPD subjects [18]. Liu et al. [15] suggested that the Fres parameter was the best IOS parameter for detecting COPD. The Fres value ≥ 17.72 Hz could detect COPD with a sensitivity of 78.9% and specificity of 93.1%. However, the cutoff values of IOS parameters in this study were derived from a comparison between COPD and healthy subjects. This may mean that the AuROC, sensitivity, and specificity may be overestimated. Further research in this area is still required as more data sets are required to look at the possible predictive value of IOS for detection of airway obstruction in chronic smoker subjects with or without COPD. Therefore, the aim of this study was to assess the diagnostic contribution of IOS to facilitate distinguishing between COPD and chronic smokers.

Materials and methods

Study procedures

This cross-sectional study was conducted during a single visit in suspected COPD subjects with a smoking history ≥ 10 pack-years who were referred by their physicians for spirometric testing at the Lung health center, Division of Pulmonary Critical Care and Allergy, Department of Internal Medicine, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand, between 15 June 2019 and 31 January 2020. IOS was measured prior to spirometry in all subjects. All tests were performed by a well-trained physical therapist. Demographic data including age, sex, co-morbidities, smoking status, symptoms of COPD (cough, phlegm, and dyspnea), age of symptom onset, duration of symptoms, and chest radiographic results were recorded. The study was approved by the Research Ethics Committee of the Faculty of Medicine, Chiang Mai University [Institutional Review Board (IRB) approval number: MED-2562-06284, date of approval: 28 May 2019 and filed under Clinical Trials Registry (Study ID: TCTR20190614005, date of approval: 14 June 2019]. Before enrollment, written informed consent was obtained from all subjects.

Subjects

One hundred and thirty-nine suspected COPD subjects were screened for eligibility for enrollment onto the study. The inclusion criteria for screening were subjects aged and onset of chronic respiratory symptoms ≥ 40 years old (at least one of symptoms including cough, phlegm, or dyspnea), with a history of smoking ≥ 10 pack-years and no history of chronic respiratory diseases other than suspected COPD. The definition of COPD was those screened subjects with post-bronchodilator (BD) less than the statistically defined fifth percentile of normal (lower limit of normal; LLN). The definition of chronic smokers was those screened subjects with normal spirometry values. The definition of healthy control subjects were subjects aged ≥ 40 years old who had no known other chronic systemic diseases (e.g. cardiovascular diseases, kidney diseases, liver diseases, neurological diseases, musculoskeletal diseases, autoimmune diseases and malignancy) and no chronic respiratory symptoms history of chronic respiratory diseases, nonsmokers or ex-smokers with a smoking history < 5 pack-years and normal spirometry values. The exclusion criteria were the subjects who had previous history of physician diagnosed-asthma, history of recurrent wheezing before age 10 year, family history positive for asthma and subjects who were unable to perform an acceptable spirometry according to the American Thoracic Society (ATS)/European Respiratory Society (ERS) 2005 standard [21]. Subjects who were unable to perform the IOS according to the standard recommended by the ERS standard [12] were also excluded. Thirty age- and gender-matched healthy subjects were also recruited onto the study.

Definition of COPD

The ATS/ERS guideline suggested that the ratio of FEV₁/forced vital capacity (FVC) varies with age [22]. Thus, using a fixed threshold criterion (FEV₁/FVC < 0.7) may result in over- or underestimation of COPD, especially in advancing or young age groups, respectively [22–25]. Therefore, the diagnosis of COPD in this study was based on the post- BD ratio of FEV₁/FVC < below LLN [22]. However, GOLD guidelines suggested that the diagnosis of COPD should be based on the post-BD ratio of FEV₁/FVC < 0.7 [1]. Therefore, the primary analysis of this study was based on LLN of FEV₁/FVC for an age-adjusted reference value. An analysis using the fixed ratio of FEV₁/FVC < 0.7 was also reported.

Spirometry

All subjects were assessed for pulmonary function using a spirometer (Vmax Encore 22, Care Fusion, Hoechberg, Germany). Pre- BD spirometry was performed according to the standards of ATS/ERS [21]. In case of the pre-BD ratio of FEV₁/FVC < LLN, the post-BD spirometry was performed 15 min after inhaling 400 µg of salbutamol from a pressurized metered-dose inhaler via a spacer device. Subjects were diagnosed as COPD based on a post-BD ratio of FEV₁/FVC < LLN which were in accordance with the ATS/ERS guideline [22]. The diagnosis of COPD according to a ratio of FEV₁/FVC < LLN was set as the reference standard in this study. Spirometric parameters were collected including FVC, FEV₁, ratio of FEV₁/FVC, and the average expired flow over the middle half (25%-75%) of the FVC maneuver (FEF25-75%). The predicted values of FVC and FEV₁ were calculated from the Global Lung Initiative (GLI) 2012 (Southeast Asian sub-group) reference equations [26].

Impulse oscillometry (IOS)

Pre-BD IOS was performed in all subjects prior to spirometry. The pulmonary resistance and reactance were measured using IOS (Master Screen IOS, Viasys GmbH, Hoechberg, Germany). Respiratory resistance (Rrs) may be largely interpreted as airway caliber [12]. The Rrs includes the resistance of the oropharynx, larynx, trachea, large and small airways, lung and chest wall tissue. Therefore, the narrower and longer airways have higher resistances due to greater frictional pressure loss as air flows through them [12]. The parameters in IOS that represent the resistance of the airways include the resistance at 5 Hz (R5), the resistance at 20 Hz (R20), and the subtraction of the R20 from R5 (R5 − R20) values. Thus, higher resistances due to greater frictional pressure loss as air flows through them was shown in narrower and longer airways [12]. Rrs is also affected by the heterogeneous distribution of resistances and reactances across the airway tree, where increasing heterogeneity increases the effective resistance at any given frequency [12].The reactance (Xrs) is the out-of-phase component of lung impedance and reflects the capacitive and inertial properties of the airways [10]. Therefore, the parameters in IOS that represent the reactance of the airways include the reactance at 5 Hz (X5), resonant frequency (Fres), and reactance area (AX). X5, the indicator of elastic recoil of the peripheral airways; Fres, indicator of the frequency as Xrs crossed zero and the elastic and inertial forces were equal in magnitude and opposite; and AX, the area of the negative reactance between 5 Hz and Fres to the zero line [10].

The subjects were asked to perform tidal breathing for 30-40 s via a mouthpiece that was connected to a loudspeaker which generates pressure oscillations composed of multiple frequencies. Subjects were asked to sit on a chair in an upright position, wore a nose clip and were directed to firmly support their cheeks with both hands. A minimum of three trials were performed in accordance with the standard recommended by ERS standard [12]. The average values from three IOS measurements were recorded. We collected the following IOS parameters: respiratory impedance values at 5 Hz (Z5), indicator of the sum of all forces which oppose the generated impulse, airway resistance (R5, R20, and R5-R20) and airway reactance (X5, Fres, and AX).

Study size calculation

The study size of the study was calculated based on data from the pilot study using MedCalc Version 18.11.6 [27] which consisted of 73 suspected COPD subjects (44 COPD and 29 chronic smokers, the ratio of sample size in negative/positive was 0.66). The AuROC of the R5 to indicate those who had COPD was 0.653. Therefore, at least 112 subjects (67 COPD and 45 chronic smokers) needed to be included in this study (power = 0.8 with statistical significance < 0.05).

Statistical analysis

Results for continuous data were expressed as mean ± standard deviation (SD) or median, interquartile range (IQR) as appropriate. Results for non-continuous data were expressed as frequencies and percentages. One way analysis of variance (ANOVA) was used to analyze differences in baseline characteristics, IOS and spirometry parameters between the three groups. Independent sample t-tests and the Mann-Whitney U Test were used to compare differences between the two groups for parametric and non-parametric data, respectively. Fisher’s exact test was used to compare the categorical data between groups. Receiver operating characteristic (ROC) curve was constructed to differentiation between COPD and chronic smokers by area under the curve (AUC) and 95% confidence interval (CI). The AuROC and 95%CI were compared among COPD and normal subjects. The AuROC and 95%CI were also compared among COPD with high and low %predicted of FEV₁ (based on median value % predicted of FEV₁ of groups) and chronic smokers. Contingency tables were made to calculate the following performance parameters: sensitivity, specificity, positive likelihood ratio (+LR) (an increase of the probability of having a disease, given a positive test result), negative likelihood ratio (-LR) (a decrease of the probability of having a disease, given a negative test result), and AUC from the various variables of IOS parameters to identify the optimum cutoff point for diagnosis of COPD. Statistical significance was accepted at a p-value < 0.05. All statistical analyses were performed using STATA version 15 (StataCorp, College Station, TX, USA).

Results

One hundred and forty-seven suspected COPD subjects were recruited onto this study. After exclusion of 30 subjects due to a history of respiratory disease other than COPD (n = 16) and unacceptable spirometry (n = 14), 117 subjects (67 COPD and 50 chronic smokers) were included in the analysis. Thirty eight healthy volunteers were also screened as a control group in this study. However, 8 of them were excluded because of unacceptable spirometry (n = 3), history of respiratory diseases (n = 2), and a smoking history ≥ 5 pack-years (n = 3). More details are shown in Figure 1.

Figure 1. Flow-chart describing the study population.

The baseline characteristics of subjects in the three groups are shown in Table 1. There were no significant differences in age, proportion of male sex, height, body weight and body mass index (BMI) among the three groups. There was no significant difference in smoking status, proportion of symptoms including cough, phlegm, and dyspnea, and age of symptom onset between COPD and chronic smokers. There was significant higher of proportion of emphysematous change in chest X-rays and duration of symptoms in COPD compared to the chronic smokers. All of suspected COPD subjects have at least one of symptoms including cough, phlegm, or dyspnea.

Table 1. Baseline characteristics of subjects. (N = 147).

Clinical characteristics	COPD (n = 67)	Chronic smokers (n = 50 )	Healthy subjects (n = 30)	p-value
Male n (%)	58 (86.6)	45 (90.0)	25 (83.3)	0.693
Age (year)	68.2 ± 8.3	69.1 ± 9.2	68.5 ± 8.1	0.869
Height (cm)	158.7 ± 7.2	159.7 ± 6.2	159.7 ± 6.8	0.631
Body weight (kg)	56.9 ± 12.9	56.6 ± 12.3	58.9 ± 9.7	0.698
Body mass index (km/m^2)	22.5 ± 4.5	22.1 ± 4.2	22.9 ± 2.6	0.672
Smoking status n (%)^a				1.000
Non-smoker	0 (0.0)	0 (0.0)	23 (76.7)
Current-smoker	4 (6.0)	3 (6.0)	0 (0.0)
Ex-smoker	63 (94.0)*	47 (94.0)*	7 (23.3)
Smoking pack-years^b	31.7 ± 24.4*	33.2 ± 20.4*	0.92 ± 1.74	0.726
Signs and symptoms (yes), n (%)
Cough^a	38 (56.7)	34 (68.0)	NA	0.252
Phlegm^a	25 (37.0)	19 (38.0)	NA	1.000
Dyspnea^a	56 (83.6)	34 (68.0)	NA	0.075
Emphysematous change in CXR (yes)^a	21 (31.3)	4 (8.0)	NA	0.003
Age of symptom onset (year)^b	66.2 ± 8.9	67.9 ± 9.8	NA	0.336
Duration of symptoms (month) (Median, IQR)^c	6 (2-12)	3 (1-12)	NA	0.025
Comorbidity n (%)
Cardiovascular disease^a	42 (62.7)*	29 (58.0)*	7 (23.3)	0.703
Metabolic disease^a	11 (16.4)*	19 (38.0)*	0 (0.0)	0.010
Neuromuscular disease^a	7 (10.4)	1 (2.0)	0 (0.0)	0.135

Note: Data are mean ± standard deviation (SD) unless otherwise stated; ^ap-value from exact test comparing COPD and chronic smokers; ^bp-value from t-test comparing COPD and chronic smokers; ^c, p-value from Mann-Whitney U test comparing COPD and chronic smokers; *, p < 0.001 compared with healthy subjects group.

Abbreviations: CXR, chest X-rays; IQR, interquartile range.

Spirometric data of the subjects are shown in Table 2. In the COPD subjects, a significant decrease in all parameters of spirometry including FEV₁, %predicted of FEV₁, FVC, %predicted of FVC, ratio of FEV₁/FVC, and FEF25-75% were observed in comparison with the chronic smokers and healthy control subjects. A significant decrease in all parameters of spirometry, except for the FEV₁ and the ratio of FEV₁/FVC, was also demonstrated in chronic smokers when compared with the healthy control subjects.

Table 2. Spirometric data of subjects.

Spirometry data (Pre-bronchodilator)	COPD (n = 67)	Chronic smokers (n = 50 )	Healthy subjects (n = 30)	p-value
FEV1 (L)^a	1.29 ± 0.44*	1.96 ± 0.49*	2.29 ± 0.56	<0.001
%predicted of FEV1^a	55.9 ± 16.6*	84.6 ± 14.3*	99.0 ± 16.4	<0.001
FVC (L)^a	2.29 ± 0.65*	2.59 ± 0.64	2.89 ± 0.58	0.039
%predicted of FVC^a	79.6 ± 17.9*	89.3 ± 16.0*	99.8 ± 12.9	0.006
FEV1/FVC (%)^a	55.9 ± 9.3*	76.2 ± 7.4	78.1 ± 5.4	<0.001
FEF25-75% (L/sec)^a	0.58 ± 0.25*	1.69 ± 0.74*	2.16 ± 0.77	<0.001

Note: Data are mean ± standard deviation (SD); ^a, p-value from ANOVA with Bonferroni adjustment compared between COPD and chronic smokers; *, p < 0.001 compared healthy subjects group from ANOVA with Bonferroni adjustment.

Abbreviations: FVC, forced vital capacity; FEV₁, forced expiratory volume in the first second; FEF25-75%, forced expiratory flow at 25-75% of FVC; L, liter.

IOS parameters of subjects are shown in Table 3. In the COPD subjects, a significant increase in parameters of IOS including Z5, R5, R20, R5-R20, Fres, and AX were recorded in comparison with the chronic smokers and healthy control subjects. A significant decrease in X5 was seen in COPD compared to chronic smokers and healthy control subjects. There were no significant differences in all parameters between chronic smokers and healthy control subjects, except for the Fres.

Table 3. IOS parameters of subjects.

IOS parameters	COPD (n = 67)	Chronic smokers (n = 50 )	Healthy subjects (n = 30)	p-value
Z5 (cmH2O/L/s)^a	5.53 ± 2.05*	3.54 ± 0.96	3.22 ± 0.78	<0.001
R5 (cmH2O/L/s)^a	4.96 ± 1.71*	3.32 ± 0.91	3.06 ± 0.72	<0.001
R20 (cmH2O/L/s)^a	3.16 ± 1.03*	2.61 ± 0.71	2.58 ± 0.64	0.002
R5 - R20 (cmH2O/L/s)^a	1.79 ± 1.05*	0.71 ± 0.47	0.48 ± 0.26	<0.001
X5 (cmH2O/L/s)^a	−2.36 ± 1.34*	−1.18 ± 0.46	−0.99 ± 0.40	<0.001
Fres (Hz)^a	23.50 ± 5.48*	16.65 ± 4.05*	13.93 ± 2.70	<0.001
AX (cmH2O/L)^a	21.14 ± 14.76*	6.36 ± 4.44	3.51 ± 2.01	<0.001

Note: Data are mean ± standard deviation (SD); ^a, p-value from ANOVA with Bonferroni adjustment compared between COPD and chronic smokers; *, p < 0.001 compared healthy subjects group from ANOVA with Bonferroni adjustment.

Abbreviations: Z5, impedance at 5 hertz; R5, resistance at 5 Hz; R20, resistance at 20 Hz, R5-R20; heterogeneity of resistance between R5 and R20; Fres, resonant frequency; X5, reactance at 5 Hz; AX, area under reactance curve between 5 Hz and resonant frequency.

The AuROC and 95%CI were compared among COPD and chronic smokers. Most of IOS parameters including AX, Fres, R5-R20, X5, Z5, and R5 demonstrated excellent overall accuracy relative to the diagnosis of COPD using the ratio of FEV₁/FVC < LLN and the ratio of FEV₁/FVC < 0.7 with an AuROC ranging from 0.80 to 0.86 (Figure 2A) and 0.83 to 0.89 (Figure 2B), respectively.

Figure 2. Receiver operating characteristic curves illustrating the performance of IOS variables in diagnosis of COPD using the ratio of FEV₁/FVC < LLN (A) and the ratio of FEV₁/FVC < 0.7 (B) between COPD and chronic smokers.

Note: AuROC from score based ROC.

The AuROC and 95%CI were also compared among COPD and normal subjects. Most of IOS parameters including AX, Fres, R5-R20, X5, Z5, and R5 demonstrated excellent overall accuracy relative to the diagnosis of COPD using the ratio of FEV₁/FVC < LLN and the ratio of FEV₁/FVC < 0.7 with an AuROC ranging from 0.86 to 0.96 (Figure 3A) and 0.87 to 0.95 (Figure 3B), respectively.

Figure 3. Receiver operating characteristic curves illustrating the performance of IOS variables in diagnosis of COPD using the ratio of FEV₁/FVC < LLN (A) and the ratio of FEV₁/FVC < 0.7 (B) between COPD and normal subjects.

Note: AuROC from score based ROC.

The sensitivity, specificity, LR+, LR-, and AUC for each cutoff point of IOS parameters for diagnosis COPD using FEV₁/FVC < LLN and FEV₁/FVC < 0.7 were comparable (Table 4). The cutoff AX ≥ 8.66 cmH₂O/L represented the highest AUC (0.79), with a sensitivity of 79.1% and specificity of 78.0% for diagnosis of COPD using the ratio of FEV₁/FVC < LLN. The cutoff AX ≥ 8.66 cmH₂O/L also represented the highest AUC (0.84), with a sensitivity of 80.6% and specificity of 86.7% for diagnosis of COPD using the ratio of FEV₁/FVC < 0.7. The second parameter that could be used for diagnosis of COPD in both criteria (FEV₁/FVC < LLN and FEV₁/FVC < 0.7) was R5-R20 ≥ 1.00 cmH₂O/L/s. The combination of AX and R5-R20 represented an AUC = 0.75, with a sensitivity of 68.7% and specificity of 82.0% and an AUC = 0.78, with a sensitivity of 66.7% and specificity of 88.9% for diagnosis of COPD using the ratio of FEV₁/FVC < LLN and the ratio of FEV₁/FVC < 0.7, respectively.

Table 4. Diagnostic performances of IOS parameters for diagnosis of COPD using the ratio of FEV₁/FVC < LLN and the ratio of FEV₁/FVC < 0.7 between COPD and chronic smokers.

IOS variable	Cutoff value	Sensitivity	Specificity	+ LR	- LR	AUC
Diagnosis of COPD using the ratio of FEV1/FVC < LLN
AX (cmH2O/L)	≥8.66	79.1	78.0	3.60	0.27	0.79
Fres (Hz)	≥19.48	70.1	78.0	3.19	0.38	0.74
R5 – R20 (cmH2O/L/s)	≥1.00	70.1	78.0	3.19	0.38	0.74
Z5 (cmH2O/L/s)	≥3.88	82.1	66.0	2.41	0.27	0.74
R5 (cmH2O/L/s)	≥3.68	79.1	68.0	2.47	0.31	0.74
X5 (cmH2O/L/s)	≤−1.33	79.1	66.0	2.33	0.32	0.73
R20 (cmH2O/L/s)	≥2.72	67.2	64.0	1.87	0.51	0.66
Combined AX ≥ 8.66 and Fres ≥ 19.48		67.2	84.0	4.20	0.39	0.76
Combined AX ≥ 8.66 and R5 – R20 ≥ 1.00		68.7	82.0	3.81	0.38	0.75
Diagnosis of COPD using the ratio of FEV1/FVC < 0.7
AX (cmH2O/L)	≥8.66	80.6	86.7	6.04	0.22	0.84
R5 – R20 (cmH2O/L/s)	≥1.00	72.2	86.7	5.42	0.32	0.79
Z5 (cmH2O/L/s)	≥3.88	83.3	73.3	3.13	0.23	0.78
R5 (cmH2O/L/s)	≥3.68	80.6	75.6	3.30	0.26	0.78
Fres (Hz)	≥19.46	72.2	82.2	4.06	0.34	0.77
X5 (cmH2O/L/s)	≤−1.33	80.6	73.3	3.02	0.27	0.77
R20 (cmH2O/L/s)	≥2.72	68.1	68.9	2.19	0.46	0.68
Combined AX ≥ 8.66 and R5 – R20 ≥ 1.00		66.7	88.9	6.00	0.38	0.78

Note: AUC from cut off based ROC.

Abbreviations: AUC, area under the curve; Z5, impedance at 5 hertz; R5, resistance at 5 Hz; R20, resistance at 20 Hz, R5-R20; heterogeneity of resistance between R5 and R20; Fres, resonant frequency; X5, reactance at 5 Hz; AX, area under reactance curve between 5 Hz and resonant frequency; + LR, positive likelihood ratio; - LR, negative likelihood ratio.

The AuROC and 95%CI were also compared among COPD with high and low %predicted of FEV₁ based on median value of group and chronic smokers. For COPD with high % predicted of FEV₁ (≥ 53.9%), most of IOS parameters including AX, Fres, R5-R20, X5, Z5, and R5 demonstrated good accuracy relative to the diagnosis of COPD using the ratio of FEV₁/FVC < LLN with an AuROC ranging from 0.72 to 0.77 (Figure 4A). For COPD with low % predicted of FEV₁ (< 53.9%), most of IOS parameters including AX, Fres, R5-R20, X5, Z5, and R5 demonstrated excellent overall accuracy relative to the diagnosis of COPD using the ratio of FEV₁/FVC < LLN with an AuROC ranging from 0.87 to 0.96 (Figure 4B).

Figure 4. Receiver operating characteristic curves illustrating the performance of IOS variables in diagnosis of COPD using the ratio of FEV₁/FVC < LLN between COPD based on median %predicted of FEV₁ (A; %predicted of FEV₁ ≥ median value of group, B; % predicted of FEV₁ < median value of group) and chronic smokers.

Note: AuROC from score based ROC.

The sensitivity, specificity, LR+, LR-, and AUC for each cutoff point of IOS parameters for diagnosis COPD using FEV₁/FVC < LLN between COPD (based on the median FEV₁) and chronic smokers were also comparable (Table 5). The diagnostic ability of IOS parameters for detection of COPD using the ratio of FEV₁/FVC < LLN was higher when compared between COPD with low % predicted of FEV₁ and chronic smokers. For COPD with low % predicted of FEV₁ (< 53.9%), the cutoff R5-R20 ≥ 1.50 cmH₂O/L/s represented the highest AUC (0.92), with a sensitivity of 87.9% and specificity of 96.0% for diagnosis of COPD. For COPD with high %predicted of FEV₁ (≥ 53.9%), the cutoff AX ≥ 8.66 cmH₂O/L represented the highest AUC (0.71), with a sensitivity of 64.7% and specificity of 78.0% for diagnosis of COPD.

Table 5. Diagnostic performances of IOS parameters for diagnosis of COPD using the ratio of FEV₁/FVC < LLN between COPD (based on the median FEV₁) and chronic smokers.

IOS variable	Cutoff value	Sensitivity	Specificity	+ LR	- LR	AUC
COPD with % predicted of FEV1 ≥ median value of group (53.9%)
AX (cmH2O/L)	≥ 8.66	64.7	78.0	2.94	0.45	0.71
X5 (cmH2O/L/s)	≤ −1.33	73.5	66.0	2.16	0.40	0.70
Z5 (cmH2O/L/s)	≥ 3.91	73.5	66.0	2.16	0.40	0.70
R5 (cmH2O/L/s)	≥ 3.68	70.6	68.0	2.21	0.43	0.69
Fres (Hz)	≥ 17.26	76.5	60.0	1.91	0.39	0.68
R5 – R20 (cmH2O/L/s)	≥ 0.84	70.6	60.0	1.76	0.65	0.65
R20 (cmH2O/L/s)	≥ 2.72	64.7	64.0	1.80	0.55	0.64
COPD with % predicted of FEV1 < median value of group (53.9%)
R5 – R20 (cmH2O/L/s)	≥ 1.50	87.9	96.0	21.97	0.13	0.92
AX (cmH2O/L)	≥ 11.84	93.9	88.0	7.83	0.07	0.91
Fres (Hz)	≥ 22.42	90.9	92.0	11.36	0.10	0.91
Z5 (cmH2O/L/s)	≥ 4.56	81.8	88.0	6.82	0.21	0.85
R5 (cmH2O/L/s)	≥ 4.16	75.8	76.0	3.16	0.32	0.76
X5 (cmH2O/L/s)	≤ −1.34	84.8	66.0	2.50	0.23	0.75
R20 (cmH2O/L/s)	≥ 2.72	69.7	64.0	1.94	0.47	0.67

Note: AUC from cut off based ROC.

Abbreviations: AUC, area under the curve; Z5, impedance at 5 hertz; R5, resistance at 5 Hz; R20, resistance at 20 Hz, R5-R20; heterogeneity of resistance between R5 and R20; Fres, resonant frequency; X5, reactance at 5 Hz; AX, area under reactance curve between 5 Hz and resonant frequency; + LR, positive likelihood ratio; - LR, negative likelihood ratio.

Discussion

COPD is characterized by an airflow limitation that is mainly attributable to the airway narrowing caused by wall remodeling and wall collapsibility during expiration due to loss of alveolar attachment and elastic recoil of the small airways [1]. Previous studies have assumed that that the IOS can provide total mechanical properties of the respiratory system [9,15]. In this study, we showed a significant increase in airway resistance and reactance in COPD and that there is a potential role for the use of IOS parameters in the diagnosis of COPD.

Respiratory resistance (Rrs) may be largely interpreted as airway caliber [12]. The narrower and longer airways have higher resistances due to greater frictional pressure loss as air flows through them [12]. Our study showed an increase in total airway resistance (R5), large airways resistance (R20) and heterogeneity of resistance (R5-R20) in COPD subjects. Additionally, the R5 and R5-R20 were significantly higher in COPD subjects in comparison to both chronic smokers and healthy control subjects. More comparative information between the results and the results from the previous studies are shown in Table 6.

Table 6. IOS parameters of subjects in our study compared to previous studies.

IOS parameters	COPD		Chronic smokers		Healthy subjects
	Our study	Previous studies	Our study	Previous studies	Our study	Previous studies
R5 (cmH2O/L/s)	4.86	3.26-4.99^[9,19,20]	3.20	3.05-3.16^[9,20]	3.06	2.75-3.36^[19,20]
R20 (cmH2O/L/s)	3.14	2.34-3.06^[9,18-20]	2.56	2.45-2.55^[9,20]	2.58	2.54-2.65^[19,20]
R5-R20(cmH2O/L/s)	1.71	0.81-1.93^[9,19,20]	0.64	0.61-1.71^[9,20]	0.48	0.20-0.71^[19,20]
X5 (cmH2O/L/s)	−2.26	−1.02- −2.95^[9,14,15,19,20]	−1.10	−0.81- −0.92^[9,20]	−0.99	−0.82- −1.02^[14,15,19,20]
Fres (Hz)	23.07	20.7-25.5^[15,18-20]	16.13	12.1-14.37^[14,20]	13.92	11.21-15.70^[14,15,19,20]
AX (cmH2O/L)	19.76	18.4-20.25^[18,20]	5.48	2.03-3.47^[9,20]	3.51	1.93-3.87^[9,20]

Abbreviations: R5, resistance at 5 Hz; R20, resistance at 20 Hz, R5-R20; heterogeneity of resistance between R5 and R20; Fres, resonant frequency; X5, reactance at 5 Hz; AX, area under reactance curve between 5 Hz and resonant frequency.

Respiratory reactance is comprised of both inertance and elastance [10,12,28,29]. More negative reactance indicates greater elastance or stiffness [12]. This typically occurs in subjects with obstructive airways diseases [12]. Additionally, the increase in AX represents the reduction in lung compliance and stiffer lung tissue associated with COPD [29]. The previous studies showed that the X5 and AX in healthy adults ranged from −0.82 to −1.02 cmH₂O/L/s [14,15,19,20] and 1.93 to 3.87 cmH₂O/L [9,20], respectively (Table 6). Our study showed a decrease in X5 and an increase in AX in subjects with COPD. Our study also demonstrated that the X5 and AX were significantly decreased and increased, respectively, in COPD subjects in comparison with both chronic smokers and healthy subjects. These results are comparable with the findings from previous findings as shown in Table 6.

Previous studies suggested that many variables including Fres, X5, R5-R20, AX of IOS, could predict COPD [14,15]. The Fres and X5 showed a high AuROC (0.90 and 0.85, respectively) for detection of COPD in a previous study [15]. Our results were comparable with the previous studies that most of IOS parameters including AX, Fres, R5-R20, X5, Z5, and R5 demonstrated excellent overall accuracy relative to the diagnosis of COPD (COPD vs. normal subjects) with an AuROC ranging from 0.86 to 0.96. Another study also suggested that the X5, R5, R5-R20, X5 and Fres readings showed a high AuROC for early detection of COPD (stage 0) [14]. However, the AUC of IOS parameters in these studies were derived from the data comparison between COPD patients and healthy normal subjects. To the best of our knowledge, our study is the first study to demonstrate the ability of IOS parameters for making the differentiation of COPD from chronic smokers. Most of the IOS parameters, including AX, Fres, R5-R20, X5, Z5, R5, could be used to diagnose COPD using the ratio of FEV₁/FVC < LLN with an AuROC ranging from 0.80 to 0.86. In general, an AuROC range from 0.7 to 0.8 and 0.8 to 0.9 are considered as acceptable and excellent diagnostic tests, respectively [30]. Therefore, our results confirm that IOS parameters demonstrated excellent overall accuracy for diagnosis of COPD.

A previous study suggested that the Fres was the best IOS parameter for detection of COPD. The Fres cutoff value of ≥ 17.72 Hz had a sensitivity of 78.9% and specificity of 93.1% for diagnosis of COPD [15]. However, this cutoff value of Fres was derived from the comparison between COPD and healthy subjects. Our study found more information about the use of IOS in the differentiation between COPD and chronic smokers. We found that AX, Fres, and R5-R20 had a good AuROC value for differential diagnosis of COPD from chronic smokers). The cutoff AX ≥ 8.66 cmH₂O/L represented the highest AUROC (0.79), with a sensitivity of 79.1% and specificity of 78.0% for diagnosis of COPD. The other parameters that could be used to diagnose COPD was Fres ≥ 19.48 Hz and R5-R20 ≥ 1.00 cmH₂O/L/s (AuROC = 0.74, sensitivity = 70.1%, and specificity = 78.0%). The combination of AX and Fres and AX and R5-R20 had an AuROC = 0.76 and 0.75 for diagnosis of COPD using FEV₁/FVC < LLN, respectively. Thus, we suggested that the use of a single parameter has adequate information for the diagnosis of COPD. We suggested that AX is the best, simple, and easy to remember. Therefore, we recommend using AX at the easy-to-remember cutoff value of AX ≥ 8.66 cmH₂O/L as the best single parameter for the diagnosis of COPD using FEV₁/FVC < LLN.

We also investigated the diagnostic performance of IOS parameters for diagnosis of COPD with high and low % predicted of FEV₁ (based on median value of group) and chronic smokers. We found that the diagnostic ability of IOS parameters (R5-R20, Fres, and AX) for detection of COPD using the ratio of FEV₁/FVC < LLN was high when compared between COPD with low % predicted of FEV₁ (< 53.9%) and chronic smokers (AUC ranged from 0.91 − 0.92). Our results were supported by the previous study indicating that IOS parameter including R5-R20, Fres, and AX can be used as an alternative method for pulmonary function assessment in patients with COPD with % predicted of FEV₁ < 50% who need inhalational glucocorticoid therapy (AUC > 0.7) [18].

Our study showed that the IOS is may be a useful tool for the diagnosis of COPD. However, there are some false positive or false negative in the diagnosis of COPD using IOS parameters in our study. Therefore, the spirometry is required for the diagnosis of COPD. The IOS cannot replace spirometry. However, the IOS can be used in subjects who have difficulties performing a forced expiration that were report about 20% in the previous studies [5,6].

Strength and limitations of this study

The strength of our study is its value as the first study that identifies the cutoff points of several IOS parameters for accurate diagnosis of COPD in smokers with suspected COPD. However, this study has some limitations. Firstly, this is single-center study. The cutoff value for diagnosis of COPD may differ in other clinical practices. Further research is needed to enroll smokers from a range of centers to add weight to the findings described in this report. Secondly, most of COPD subjects have low FEV₁. Only 15 (22.4%) of COPD subjects have %predicted of FEV₁ greater than 70%. Therefore, there is limited sample size for analyzing between subjects with low FEV₁ and subjects with preserved FEV₁. The comparison between subjects with low FEV₁ and subjects with preserved FEV₁ should be mentioned in the future study. Thirdly, bronchodilation is only performed in subjects with the pre-BD ratio of FEV₁/FVC < LLN. This means that volume responders are easily missed. Therefore, pre- and post-BD spirometry should be performed in all suspected COPD subjects in the future study. Lastly, the airway resistance and reactance was measured at a tidal breathing. Further study should separately measure at the inspiratory and expiratory for contributions of airway resistance and reactance.

Conclusion

Our study indicates that the majority of IOS parameters, including AX, Fres, R5-R20, X5, Z5, and R5, result in an excellent overall level of accuracy for differential diagnosis of COPD from chronic smoker. The AX ≥ 8.66 cmH₂O/L has ability for diagnosing COPD. The diagnostic ability of IOS parameters for detection of COPD was high when compared between COPD with low % predicted of FEV₁ and chronic smokers. These findings indicate that IOS is a useful tool for diagnosis of COPD especially in the subjects who cannot perform spirometry.

Author contributions

Conceptualization, W.C., S.N., C.L., and C.P.; Methodology, W.C., S.N., C.L.; and C.P. Software, W.C.; Validation, W.C., S.N., C.L., and C.P.; Formal Analysis, W.C., S.N.; Investigation, W.C., C.L., and C.P.; Resources, W.C., C.L., and C.P.; Data Curation, W.C. and C.P.; Writing – Original Draft Preparation, W.C.; Writing – Review & Editing, W.C., S.N., C.L., and C.P.; Visualization, W.C., S.N., C.L., and C.P.; Supervision, S.N., C.L., and C.P.; Project Administration, W.C. and C.P.; Funding Acquisition, W.C., S.N., and C.P. All authors have read and agreed to the published version of the manuscript.

Abbreviations
ANOVA	=	One way analysis of variance
ATS	=	American Thoracic Society
AUC	=	Area under the curve
AuROC	=	Area under the receiver operating characteristic curve
AX	=	Area under reactance curve between 5 Hz and Fres
BD	=	Bronchodilator
BMI	=	Body mass index
CI	=	Confidence interval
COPD	=	Chronic obstructive pulmonary disease
ERS	=	European Respiratory Society
FEF25-75%	=	Average expired flow over the middle half of the FVC maneuver
FEV1	=	Forced expiratory volume in the first second
FVC	=	Forced vital capacity
Fres	=	Resonant frequency
GLI	=	Global Lung Initiative
GOLD	=	Global Initiative for Chronic Obstructive Lung Disease
IOS	=	Impulse oscillometry
IQR	=	Interquartile range
LR	=	Likelihood ratio
R5	=	Resistance at 5 Hz
R20	=	Resistance at 20 Hz
R5-R20	=	Subtraction of the R20 from R5
ROC	=	Receiver operating characteristic
Rrs	=	Respiratory resistance
SD	=	Standard deviation
WHO	=	World Health Organization
X5	=	Reactance at 5 Hz

Acknowledgments

The authors would like to thank all subjects who kindly participated in this study. The authors acknowledge the physicians and nurses of the Division of Pulmonary, Critical Care and Allergy, Department of Internal Medicine, Faculty of Medicine, Chiang Mai University for their contribution to this trial. Finally, we acknowledge the Faculty of Medicine Graduate Student Scholarship, Chiang Mai University for the supporting.

Disclosure statement

The authors have no conflicts of interest in connection with this work.

Additional information

Funding

This study is funded by the Faculty of Medicine, Chiang Mai University Research Fund under grant No. 037/2563.